Embodiments of the invention relate to an image processing method and an image processing apparatus for correcting skin color in an input image, a digital photographing apparatus using the image processing apparatus, and a computer readable storage medium for executing the image processing method.
Since digital photographing apparatuses and image processing apparatuses are being widely distributed, demand for showing beautiful faces in captured images is increasing. Accordingly, a study on a face correction technique has been actively conducted for brightening a color tone of skin, removing freckles from the face, and thereby showing a clear face. A user may correct the captured face to look beautiful by simply using the face correction technique.
The function of removing freckles or wrinkles in the face correction function completely removes or lightens the freckles or wrinkles included in the face. The technique for removing freckles may include, for example, Gaussian smoothing by using a Gaussian low pass filter and bilateral filtering.
FIG. 1 illustrates images processed by using a Gaussian low pass filter and a bilateral filter.
In skin correction using the Gaussian low pass filter, natural skin correction may not be expected. The process using the Gaussian low pass filter provides an effect wherein pixels that are determined as noise corresponding to freckles or wrinkles are appropriately mixed with surrounding pixels and thus it looks like the freckles or wrinkles are removed. Referring to a second image 120 of FIG. 1, freckles are blurred and widely spread compared with those of an input image 110. Also, in the process of using the Gaussian low pass filter, edges are blurred. Thus, in order to solve problems regarding blurred edges, other complicated processes may be performed after the filtering process using the Gaussian low pass filter. However, even if other additional processes are performed, blurred edges are still generated as shown in image 220 of FIG. 2 in which the process using the Gaussian low pass filter and other additional process are performed on an input image 210. Accordingly, if a blur is generated in the edges, a captured image may look like it is out of focus.
In addition, in the process of using the bilateral filter, the problem that edge components are blurred is less than other filters, but the edge blurring still exists. Also, when there is a large number of freckles, an unnatural image in which edges corresponding to outline of eyes, nose, and mouth and wrinkles may be obtained as illustrated in a third image 130 of FIG. 1 due to an effect in which the freckles are likely gathered together.
FIG. 3 illustrates a convolution operation for a general filtering process.
The general filtering process is performed by using center pixels 322, 324, 326, and 328 and adjacent surrounding pixels with respect to each pixel of an input image 310, as illustrated in FIG. 3. However, freckles have a larger size than general image noise and the difference in brightness between freckles and adjacent surrounding pixels is not great. Accordingly, the convolution operation by using the adjacent surrounding pixels is not effective in removing freckles in consideration of such property of freckles.
Related arts regarding skin color correction are as follows. U.S. 2002-160421, which is related to a method and system for enhancing portrait images, discloses that properties of regions such as eyes, nose, mouth, and skin are identified from a face image and the regions are processed by using appropriate filters, thereby solving problems regarding blurred edge. Japanese Patent Document No. 2005-162836 discloses that a face region is accurately extracted and the face is naturally and effectively corrected. In addition, related arts regarding face correction include Japanese Patent Document No. 1998-053026 and Japanese Patent Document No. 2002-345643. The details thereof are referred in this specification.